github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15507] LangChain: chatollama invocation fails when messages contain only SystemMessage/AIMessage without any HumanMessage #9910

New Issue
Open
opened 2026-04-12 22:45:32 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 22:45:32 -05:00
Owner
Copy Link

Originally created by @Bkbest on GitHub (Apr 11, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15507

What is the issue?

When invoking the Ollama LLM via ChatOllama with a message list that contains only SystemMessage(s)/AIMessage(s) and no HumanMessage, the invocation fails with an error. This is problematic when using message summarization libraries like langmem which can sometimes replace all messages with a single SystemMessage during the summarization process.

Steps to Reproduce:

  1. Create a ChatOllama model instance
  2. Prepare a message list containing only SystemMessage(s) - for example, after a summarization process collapses all previous messages into a single system message
  3. Attempt to invoke the model with: model.ainvoke(messages)

Model being used: minimax-m2.7:cloud (ollama pro)

Code I have:

`async def llm_with_tools(state: State, runtime: Runtime):

"""
Processes messages using LLM with tools when required.

Args:
    state: Current state containing messages and tool requirement
    
Returns:
    Dict containing updated messages
"""
info = runtime.execution_info
if info.node_attempt > 1:
    print("sleeping for 60 seconds before retrying.")
    await asyncio.sleep(60)     
# Create the prompt template with system prompt and messages
agent_description = AGENT_DESCRIPTION
prompt = PromptTemplate.from_template(agent_description)
system_message = [SystemMessage(content=prompt.format(current_date=state["current_date"], skills_description=state["skills_description"]))]

#add systemmessage to the beginning of the messages to be summarized so that it is included in the summary.
state["messages"] = system_message + state["messages"]

summarization_result = await asummarize_messages(
    state["messages"],
    running_summary=state.get("summary"),
    model=llm,
    max_tokens=10000,
    max_tokens_before_summary=5000,
    max_summary_tokens=1000,
    initial_summary_prompt=DEFAULT_INITIAL_SUMMARY_PROMPT,
    existing_summary_prompt=DEFAULT_EXISTING_SUMMARY_PROMPT,
)
messags_after_summarization = summarization_result.messages

# If there are exactly two system messages after summarization, convert the second one to an AIMessage
system_message_indices = [i for i, msg in enumerate(messags_after_summarization) if isinstance(msg, SystemMessage)]
if len(system_message_indices) == 2:
    print(system_message_indices)
    second_system_idx = system_message_indices[1]
    messags_after_summarization[second_system_idx] = AIMessage(content=messags_after_summarization[second_system_idx].content)
    # Insert a HumanMessage after the second system message using the first message from state
    # human_msg = HumanMessage(content="Look at the summary and the conversation below and decide what to do next.")
    # messags_after_summarization.insert(second_system_idx, human_msg)
    print(messags_after_summarization)
await asyncio.sleep(10)
response = await llm_tools.ainvoke(messags_after_summarization)
state_update = {"messages": [response]}
if summarization_result.running_summary:
    state_update["summary"] = summarization_result.running_summary

# Return the response as a message
return state_update`

Full Code: https://github.com/Bkbest/basic_deep_agent

Expected Behavior:

The model should be able to process message list even if it does not contain a HumanMessage.

Current Workaround:
Insert a dummy HumanMessage before the system message when this scenario occurs

Versions I am using:

Name: langchain-ollama
Version: 1.0.1

Name: langchain
Version: 1.2.10

Name: langmem
Version: 0.0.30

Relevant log output

2026-04-11T20:20:44.443753Z [info     ] Retrying task llm_with_tools after 1.15 seconds (attempt 1) after ResponseError Service Temporarily Unavailable (status code: 503) [langgraph.pregel._retry] api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca graph_id=agent langgraph_api_version=0.7.56 request_id=bd85cd81-e24d-4e35-adf2-a9eecd822482 run_attempt=1 run_id=019d7e32-29ee-7103-9a76-8df5d8300795 thread_id=2a49f20a-8cf8-4530-aa65-232198cd0833 thread_name=MainThread
Traceback (most recent call last):
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\pregel\_retry.py", line 211, in arun_with_retry
    return await task.proc.ainvoke(task.input, config)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\_internal\_runnable.py", line 705, in ainvoke
    input = await asyncio.create_task(
            ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\_internal\_runnable.py", line 473, in ainvoke
    ret = await self.afunc(*args, **kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\AI\basic-agent\basic_deep_agent\src\AI_Nodes\nodes.py", line 93, in llm_with_tools
    response = await llm_tools.ainvoke(messags_after_summarization)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\runnables\base.py", line 5708, in ainvoke
    return await self.bound.ainvoke(
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 425, in ainvoke
    llm_result = await self.agenerate_prompt(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1134, in agenerate_prompt
    return await self.agenerate(
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1092, in agenerate
    raise exceptions[0]
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1318, in _agenerate_with_cache
    async for chunk in self._astream(messages, stop=stop, **kwargs):
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 1193, in _astream
    async for chunk in self._aiterate_over_stream(messages, stop, **kwargs):
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 1131, in _aiterate_over_stream
    async for stream_resp in self._acreate_chat_stream(messages, stop, **kwargs):
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 937, in _acreate_chat_stream
    async for part in await self._async_client.chat(**chat_params):
  File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\ollama\_client.py", line 741, in inner
    raise ResponseError(e.response.text, e.response.status_code) from None
ollama._types.ResponseError: Service Temporarily Unavailable (status code: 503)
During task with name 'llm_with_tools' and id 'fa2904e7-60e4-5492-37ff-aec55365e083'
sleeping for 60 seconds before retrying.

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.17.1

Originally created by @Bkbest on GitHub (Apr 11, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15507 ### What is the issue? When invoking the Ollama LLM via ChatOllama with a message list that contains only SystemMessage(s)/AIMessage(s) and no HumanMessage, the invocation fails with an error. This is problematic when using message summarization libraries like langmem which can sometimes replace all messages with a single SystemMessage during the summarization process. Steps to Reproduce: 1. Create a ChatOllama model instance 2. Prepare a message list containing only SystemMessage(s) - for example, after a summarization process collapses all previous messages into a single system message 3. Attempt to invoke the model with: model.ainvoke(messages) Model being used: minimax-m2.7:cloud (ollama pro) Code I have: `async def llm_with_tools(state: State, runtime: Runtime): """ Processes messages using LLM with tools when required. Args: state: Current state containing messages and tool requirement Returns: Dict containing updated messages """ info = runtime.execution_info if info.node_attempt > 1: print("sleeping for 60 seconds before retrying.") await asyncio.sleep(60) # Create the prompt template with system prompt and messages agent_description = AGENT_DESCRIPTION prompt = PromptTemplate.from_template(agent_description) system_message = [SystemMessage(content=prompt.format(current_date=state["current_date"], skills_description=state["skills_description"]))] #add systemmessage to the beginning of the messages to be summarized so that it is included in the summary. state["messages"] = system_message + state["messages"] summarization_result = await asummarize_messages( state["messages"], running_summary=state.get("summary"), model=llm, max_tokens=10000, max_tokens_before_summary=5000, max_summary_tokens=1000, initial_summary_prompt=DEFAULT_INITIAL_SUMMARY_PROMPT, existing_summary_prompt=DEFAULT_EXISTING_SUMMARY_PROMPT, ) messags_after_summarization = summarization_result.messages # If there are exactly two system messages after summarization, convert the second one to an AIMessage system_message_indices = [i for i, msg in enumerate(messags_after_summarization) if isinstance(msg, SystemMessage)] if len(system_message_indices) == 2: print(system_message_indices) second_system_idx = system_message_indices[1] messags_after_summarization[second_system_idx] = AIMessage(content=messags_after_summarization[second_system_idx].content) # Insert a HumanMessage after the second system message using the first message from state # human_msg = HumanMessage(content="Look at the summary and the conversation below and decide what to do next.") # messags_after_summarization.insert(second_system_idx, human_msg) print(messags_after_summarization) await asyncio.sleep(10) response = await llm_tools.ainvoke(messags_after_summarization) state_update = {"messages": [response]} if summarization_result.running_summary: state_update["summary"] = summarization_result.running_summary # Return the response as a message return state_update` Full Code: https://github.com/Bkbest/basic_deep_agent Expected Behavior: The model should be able to process message list even if it does not contain a HumanMessage. Current Workaround: Insert a dummy HumanMessage before the system message when this scenario occurs **Versions I am using:** Name: langchain-ollama Version: 1.0.1 Name: langchain Version: 1.2.10 Name: langmem Version: 0.0.30 ### Relevant log output ```shell 2026-04-11T20:20:44.443753Z [info ] Retrying task llm_with_tools after 1.15 seconds (attempt 1) after ResponseError Service Temporarily Unavailable (status code: 503) [langgraph.pregel._retry] api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca graph_id=agent langgraph_api_version=0.7.56 request_id=bd85cd81-e24d-4e35-adf2-a9eecd822482 run_attempt=1 run_id=019d7e32-29ee-7103-9a76-8df5d8300795 thread_id=2a49f20a-8cf8-4530-aa65-232198cd0833 thread_name=MainThread Traceback (most recent call last): File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\pregel\_retry.py", line 211, in arun_with_retry return await task.proc.ainvoke(task.input, config) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\_internal\_runnable.py", line 705, in ainvoke input = await asyncio.create_task( ^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langgraph\_internal\_runnable.py", line 473, in ainvoke ret = await self.afunc(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\AI\basic-agent\basic_deep_agent\src\AI_Nodes\nodes.py", line 93, in llm_with_tools response = await llm_tools.ainvoke(messags_after_summarization) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\runnables\base.py", line 5708, in ainvoke return await self.bound.ainvoke( ^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 425, in ainvoke llm_result = await self.agenerate_prompt( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1134, in agenerate_prompt return await self.agenerate( ^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1092, in agenerate raise exceptions[0] File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_core\language_models\chat_models.py", line 1318, in _agenerate_with_cache async for chunk in self._astream(messages, stop=stop, **kwargs): File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 1193, in _astream async for chunk in self._aiterate_over_stream(messages, stop, **kwargs): File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 1131, in _aiterate_over_stream async for stream_resp in self._acreate_chat_stream(messages, stop, **kwargs): File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\langchain_ollama\chat_models.py", line 937, in _acreate_chat_stream async for part in await self._async_client.chat(**chat_params): File "C:\Users\aravi\AppData\Local\Programs\Python\Python312\Lib\site-packages\ollama\_client.py", line 741, in inner raise ResponseError(e.response.text, e.response.status_code) from None ollama._types.ResponseError: Service Temporarily Unavailable (status code: 503) During task with name 'llm_with_tools' and id 'fa2904e7-60e4-5492-37ff-aec55365e083' sleeping for 60 seconds before retrying. ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.17.1
GiteaMirror added the bug label 2026-04-12 22:45:32 -05:00
GiteaMirror commented 2026-04-12 22:45:36 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Apr 12, 2026):

$ curl localhost:11434/v1/chat/completions -d '{
    "model":"minimax-m2.7:cloud",
    "messages":[{"role":"system","content":"this is a single system message"}],
    "stream":false
  }'
{"error":"Internal Server Error (ref: 1bd5e687-9e90-42aa-b772-13c5674cb27b)"}
<!-- gh-comment-id:4232751936 --> @rick-github commented on GitHub (Apr 12, 2026): ```console $ curl localhost:11434/v1/chat/completions -d '{ "model":"minimax-m2.7:cloud", "messages":[{"role":"system","content":"this is a single system message"}], "stream":false }' {"error":"Internal Server Error (ref: 1bd5e687-9e90-42aa-b772-13c5674cb27b)"}
GiteaMirror commented 2026-04-12 22:45:36 -05:00
Author
Owner
Copy Link

@Bkbest commented on GitHub (Apr 13, 2026):

yes, even If I add an AI messages, it would fail. Only a HumanMessage will result in successful call.

<!-- gh-comment-id:4233569007 --> @Bkbest commented on GitHub (Apr 13, 2026): yes, even If I add an AI messages, it would fail. Only a HumanMessage will result in successful call.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#9910
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?